Expanding Online Chat Communications Based on Chat Context

1. A method performed by one or more computing devices, the method comprising: establishing a first chat communication group for exchanging chat messages between a plurality of client devices; inputting a subset of the chat messages within the first chat communication group to a trained machine learning (“ML”) model to generate a chat summary of the subset of the chat messages; determining a relevant user of the first chat communication group based on the chat summary; providing a recommendation for inviting the relevant user to a second chat communication group; establishing the second chat communication group; and presenting at least one of the chat summary or a message of the first chat communication group in the second chat communication group in response to the relevant user joining the second chat communication group.

2. The method of claim 1, wherein the relevant user is determined by: matching keywords associated with the chat summary with a user record in a data store; determining a number of matches of the keywords with the user record; and determining a user associated with the user record as the relevant user based on the number of matches exceeding a threshold value.

3. The method of claim 1, further comprising: generating a second chat summary of a second subset of the chat messages within the second chat communication group; determining a second relevant user of the second chat communication group based on the second chat summary; providing a recommendation for inviting the second relevant user to a third chat communication group; establishing the third chat communication group; and presenting the second chat summary in the third chat communication group in response to the second relevant user joining the third chat communication group.

4. The method of claim 1, wherein providing the recommendation for inviting the relevant user to the second chat communication group comprises presenting the recommendation in a user interface to a client computing device associated with a host of the first chat communication group, and further comprising: receiving an approval from the client computing device associated with the host of the second chat communication group; and admitting the relevant user to the second chat communication group.

5. The method of claim 1, wherein the trained ML model is trained using training data samples, wherein a training data sample comprises a group of chat messages and a chat summary generated by a second machine learning model with the group of chat messages as input.

6. The method of claim 5, wherein parameters of the trained ML model are quantized to reduce a number of bits used to represent the parameters and wherein the chat summary of the chat messages is generated by applying the trained ML model with quantized parameters.

7. The method of claim 6, wherein applying the trained ML model to the chat messages comprises: generating input embeddings for the subset of the chat messages; quantizing the input embeddings to reduce a number of bits used to represent the input embeddings; providing the quantized input embeddings to the trained ML model to generate an intermediate vector by a first layer of the trained ML model; quantizing the intermediate vector to reduce a number of bits used to represent the intermediate vector; and providing the quantized intermediate vector to a second layer of the trained ML model to generate the chat summary.

8. A system comprising: a non-transitory computer-readable medium; a communications interface; and a processor communicatively coupled to the non-transitory computer-readable medium and the communications interface, the processor configured to execute processor-executable instructions stored in the non-transitory computer-readable medium to: establish a first chat communication group for exchanging chat messages between a plurality of client devices; input a subset of the chat messages within the first chat communication group to a trained machine learning model to generate a chat summary of the subset of the chat messages; determine a relevant user of the first chat communication group based on the chat summary; provide a recommendation for inviting the relevant user to a second chat communication group; establish the second chat communication group; and present at least one of the chat summary or a message of the first chat communication group in the second chat communication group in response to the relevant user joining the second chat communication group.

9. The system of claim 8, wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to: match keywords associated with the chat summary with a user record in a data store; determine a number of matches of the keywords with the user record; and determine a user associated with the user record as the relevant user based on the number of matches exceeding a threshold value.

10. The system of claim 8, wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to: generate a second chat summary of a second subset of the chat messages within the second chat communication group; determine a second relevant user of the second chat communication group based on the second chat summary; provide a recommendation for inviting the second relevant user to a third chat communication group; establish the third chat communication group; and present the second chat summary in the third chat communication group in response to the second relevant user joining the third chat communication group.

11. The system of claim 8, wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to: present the recommendation in a user interface to a client computing device associated with a host of the first chat communication group; receive an approval from the client computing device associated with the host of the second chat communication group; and admit the relevant user to the second chat communication group.

12. The system of claim 8, wherein the trained ML model is trained using training data samples, wherein a training data sample comprises a group of chat messages and a chat summary generated by a second machine learning model with the group of chat messages as input.

13. The system of claim 12, wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to quantize parameters of the trained ML model to reduce a number of bits used to represent the parameters and apply the trained ML model with quantized parameters to generate the chat summary.

14. The system of claim 13, wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to: generate input embeddings for the subset of the chat messages; quantize the input embeddings to reduce a number of bits used to represent the input embeddings; provide the quantized input embeddings to the trained ML model to generate an intermediate vector by a first layer of the trained ML model; quantize the intermediate vector to reduce a number of bits used to represent the intermediate vector; and provide the quantized intermediate vector to a second layer of the trained ML model to generate the chat summary.

15. A non-transitory computer-readable medium comprising processor-executable instructions configured to cause one or more processors to: establish a first chat communication group for exchanging chat messages between a plurality of client devices; input a subset of the chat messages within the first chat communication group to a trained machine learning model to generate a chat summary of the subset of the chat messages; determine a relevant user of the first chat communication group based on the chat summary; provide a recommendation for inviting the relevant user to a second chat communication group; establish the second chat communication group; and present at least one of the chat summary or a message of the first chat communication group in the second chat communication group in response to the relevant user joining the second chat communication group.

16. The non-transitory computer-readable medium of claim 15, further comprising processor-executable instructions configured to cause the one or more processors to: match keywords associated with the chat summary with a user record in a data store; determine a number of matches of the keywords with the user record; and determine a user associated with the user record as the relevant user based on the number of matches exceeding a threshold value.

17. The non-transitory computer-readable medium of claim 15, further comprising processor-executable instructions configured to cause the one or more processors to: present the recommendation in a user interface to a client computing device associated with a host of the first chat communication group; receive an approval from the client computing device associated with the host of the second chat communication group; and admit the relevant user to the second chat communication group.